Personalized spaced repetition now possible

Everyone who tried spaced repetition for long enough knows that its downside in having to create the flashcards. Learning with flashcards is fun but when we have to create them, it requires context switch and often, we memorize their contents while creating them.

Sharing and reusing flashcards saves time, but it lacks the personal touch, and lacks the right context. And crunching other people's flashcards is also quite boring because we might simply not need to learn all the same things.

Thanks to the LLMs we could now generate personal flashcards on the go, based on everyday queries to LLMs and adapt them to our needs. Wrap them to the best context to match our pre-understanding of the concepts.

Im thinking of the following flow (also sketched in a diagram)

1. I use my LLM chatbot as a companion when researching topics and ask it classical questions during learning.
2. The app recognizes concepts I struggle with and generates flashcards for me. Or I can tell the app to help me memorize the concept.
3. The app analyzes the concept with respect to my previous questions. It generates a flashcard with the right context and difficulty level.
4. The app schedules repetition of the concept for me and adapts the repetition based on my performance.
5. Over time, the app understands the right contexts and learning patterns for me and helps me memorize more effectively.

I dreamed of this many many years ago, when LLMs were not a thing. At the time I did my research on this, the best models could only summarize text. Now we can finally generate flashcards on the fly with little human content creation.

I will dig deeper into this topic and share more insights in the future posts.